INADVERTENT INFLOW

When it is known for certain that the solution must entail outflow over the whole of a particular boundary-condition 'patch', users may neglect to ascribe values to the values of fluid properties at that region, on the grounds that they cannot be influential.

However, even though the final solution entails outflow, inflow MAY occur on the road to the solution, as the pressure field adjusts itself gradually.

Since fluid properties which have not been set will be taken as zero, the consequence is that fluid at the absolute zero of temperature (for example) may be treated as being drawn into the integration domain, with dire effects on the subsequent course of the solution.

Sometimes divergence results from such effects; and sometimes a converged solution is arrived at, with inflow, which differs from the one which corresponds to the situation ostensibly being simulated.

Of course, the belief that the patch should exhibit only outflow may be wrong. The patch may comprise several cells, through SOME of which fluids flows out while through others it flows in.

One satisfactory preventive measure is to ascribe physically plausible values of the fluid properties even at outlet-boundary- condition patches; then transitory inflow does no damage.

Another is so to prescribe the scalar properties of the incoming fluid (but NEVER the velocity variables) as to make no difference to what prevails in the cell; this is effected by way of the command:

COVAL (patchname,variable,ONLYMS,SAME)

This is indeed highly probable if the third argument of COVAL for P1 is large; for then small differences of pressure resulting from the varying dynamic head of the fluid can easily give rise to some cells being above and others below the fourth-argument value of the command.

If this is the case, the above mentioned use of SAME in the COVAL commands for other variables will NOT be satisfactory: it becomes essential to insert values which are properly representative of the fluid which is drawn in through the patch.

Should it be desired to eliminate this effect of the variations of dynamic head (which can cause non-convergence), a smaller COVAL coefficient for pressure is required. The magnitude of this pressure coefficient, Cp, is estimated by the following argument:

• The mass flow rate per unit area, u, is equal to Cp*(Vp - p), where Vp is the COVAL value and p is the in-cell pressure. Hence:

  rho*vel = Cp*(Vp - p)

• It is required that the difference in pressure across the outlet greatly exceeds the dynamic-head variations, ie:

  (Vp - p) >> rho*vel**2

• Eliminating (Vp - p) from the two equations leads to the required condition on Cp, namely:

  Cp << 1/vel

Here it may be remarked that one may easily deceive oneself by giving the patch a name such a OUTLET, EXIT, etc; although in fact the name can have no influence on whether fluid comes in or out, thinking may be influenced by a name which was too lightly selected. Even patch names should therefore be chosen with due care.

wbs